Areas of Expertise


Achievement Gaps
Literature Library


Why Do the Achievement Gaps Exist?

  1. Explanations

    1. School Factors

      1. Racial Segregation in Schools

Researchers Link School Segregation to the Racial Gap in Achievement

A study of Texas schools finds that having more black classmates slows the achievement gains of other black elementary school children, particularly those with above-average test scores. This "peer effect" does not affect white students, suggesting that some of the racial gap in achievement is due to segregation.

This reports some of the ideas and findings from the following source:

Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2002).New evidence about Brown v. Board of Education: The complex effects of school racial composition on achievement.National Bureau of Economic Research, Working Paper 8741. Pp. 22-29

To see other reports that originated from this same citation, go to the bibliography.

What effect did desegregation have on student achievement? Researchers Eric A. Hanushek, John F. Kain, and Steven G. Rivkin look at the effect of the race of classmates on changes in test scores for three cohorts of Texas fourth graders as they moved through elementary school. Their basic findings are presented in Table 1.

Table 1. Effects of Peer Racial Composition on Achievement Gains

(*=statistically insignificant)

Percent Black effect on:

Achievement growth student and school-by-grade fixed effects teacher and school characteristics district-by-year fixed effects

Black students





White students





Hispanic students





Racial Composition Affects the Achievement Gap

Table 1 shows how a change in the percentage of blacks in a school affects changes in achievement (expressed in standard deviations of test scores). Controlling for student and school-by-grade fixed effects, for example, shows that a 10 percentage point increase in the proportion of blacks reduces achievement growth by 0.3 standard deviations.

The authors say that adding in controls for teacher and school characteristics (class size, teacher experience, and student turnover) reduces this effect by nearly 50% (from -0.30 to -0.21). Controlling for district-by-year effects (changes in leadership, curriculum, and so on) has little effect on the results, suggesting "that the results are not being driven by an unobserved change in school quality that leads to changes in racial composition," the authors write (p. 23).

Table 1 shows no statistically reliable effect of the percentage of black classmates on the achievement gains of white and Hispanic students, which suggests that this peer effect is not due to "unobserved differences in teacher or school quality" since these would affect all groups equally (p. 24). The authors also find no effect of the percent ofHispanic students on the achievement gains of any group.

The Effect of Peers Varies by Student Achievement

Hanushek and his coauthors say that teachers may "lower expectations" as the proportion of black students rises or that some blacks may "discourage others from excelling academically" (p. 24). This suggests that the effect of black peers might be greater on students with higher initial test scores.

Table 2. Effects of Peer Racial Composition by Achievement Level

(*=statistically insignificant)

Percent Black effect on:

Bottom Quartile

Second Quartile

Third Quartile

Top Quartile

Black students





White students





The results shown in Table 2 "support the view that higher-achieving blacks are much more sensitive to school racial composition," the authors write (p. 25). Table 2 shows no statistically reliable effect of percentage black on whites or black students in grades 5&–7 with test scores below the mean. The effect of black peers is significant and increases for black students in the third and fourth quartiles.

District History and Racial Exposure Do Not Matter

Community resistance to integration may make an impact on its effect. Without information on desegregation by district, Hanushek and his coauthors look at the increase in exposure to black peers by district from 1968 to 1992. They find no evidence that the rate of integration alters the effects of racial composition on achievement.

It also is possible that the effect of peers varies by the level of racial exposure. The authors compare districts having few and many minority students and find no effect of the percentage of black students (0&–33, 34&–66, and 67&–100) on the relationship between racial composition and achievement gains.

Desegregation Would Help Close the Achievement Gap

Analysis of this large data set from Texas "supports the view that school proportion black has a negative effect on mathematics achievement growth for blacks that is concentrated in the upper half of the ability distribution," the authors conclude. "What is particularly important is that this effect does not appear to be driven by school quality differences, achievement differences of classmates, or even the specific distribution of ability within the school" (p. 28).

Since blacks still are more likely to have black classmates than whites, Hanushek and his coauthors estimate that "equalizing the black distribution" of students in Texas elementary schools would reduce the 7th-grade racial achievement gap by 25% (p. 28).

Research Design:

This study measures peer influence on student achievement by looking at changes in the racial composition and test scores in Texas schools.

A matched panel data set of school operations constructed by the UTD Texas Schools Project was used for the analysis. This data set tracks three successive cohorts of students, with the first beginning third grade in 1992. The authors use data for grades 4&–6 for one cohort and grades 4&–7 for the other two. They exclude the relatively small numbers of Asian and Native American students from their sample.

The authors use a value-added model of achievement, where changes in a student's test scores (standardized mathematics results from the Texas Assessment of Academic Skills) from one grade to the next is a function of:

  • peer behavior , which is itself assumed to be a linear function of classmate attributes such as proportion black, proportion Hispanic, average ability (measured by past peer achievement), and average family income
  • family background , such as race and socioeconomic status (eligibility for free lunch or Title I) and
  • school characteristics , such as average class size and teacher quality (years of experience and highest degree earned)

Using a value-added model circumvents the problem of omitted historical variables, the authors say, since only two observations on each student are required.

However, problems with measuring school quality, individual ability, and neighborhood and school choice by parents remain. Each of these can show up as peer influence if left uncontrolled. The authors attempt to eliminate this potential bias by breaking the error term of their model into time invariant:

  • individual effects on achievement, such as differences in ability, child-rearing practices, financial inputs, motivational influences, and parental attitudes toward schools and peers
  • school effects on achievement, such as stable differences in curriculum and teacher quality
  • school-by-grade effects on achievement, such as changes in a school's basic curriculum or the introduction of specialized programs
  • grade-by-year differences in the testing regime
  • school-by-grade effects that vary from cohort to cohort, such as changes in teacher quality or white flight
  • district-by-year effects on achievement, such as changes in hiring practices and pay, teacher and principal assignments to schools, attendance boundaries and placement rules, and desegregation plans

These "fixed effects account for the primary systematic but unobserved differences in students and schools," the authors write (p. 13). A potential correlate of racial composition that remains is family relocation in response to anticipated problems. However, the authors say that such costly decisions should lag behind changes in peer, teacher, or school characteristics.

This study was supported by the Spencer Foundation, the Mellon Foundation, the Smith Richardson Foundation, and the Packard Humanities Institute.


Copyright © 2007 Learning Point Associates. All rights reserved.